Experience layer
Patient and clinician surfaces
Web, mobile, tablet, and clinician dashboards keep records, diaries, care tasks, and consent accessible.
Technology
A secure architecture for lifelong patient intelligence.
MyRekod is designed around interoperable records, adapter-based integrations, zero-trust access, governed AI, and visible patient consent.
System layers
The platform is easier to trust when every layer has a clear job.
The architecture separates user experience, access policy, clinical services, data storage, AI, and governance so the system can scale without becoming opaque.
Access layer
Gateway and policy boundary
Identity, request routing, rate limits, authorization policies, and audit context are handled before core services.
Service layer
Clinical services
Records, diary, vitals, medication, care pathways, notifications, marketplace, and clinician collaboration run as bounded services.
Data layer
Health data foundation
Structured records, time-series vitals, imaging references, documents, terminology, and searchable context are encrypted and governed.
Model layer
AI intelligence layer
Risk screening, symptom extraction, vitals analysis, imaging workflows, recommendations, and patient explanations are served through governed endpoints.
Trust layer
Security and governance
Consent, encryption, access review, model cards, clinical validation, drift monitoring, and audit proofs run across every layer.
Integration rule
Nothing connects directly to the clinical core.
Hospitals, devices, labs, pharmacies, insurers, payment channels, imaging systems, and terminology sources pass through adapters. The adapter normalizes the payload, validates it, logs it, and then publishes it into the correct service.
PHI stays off-chainEvery access is auditableThird parties use adaptersFHIR-aligned exchangeClinical AI requires validationModel drift is monitored
Security posture
Security is not a feature at the end. It is the operating boundary.
Identity, encryption, consent, policy checks, audit logs, and PHI boundaries are applied consistently across client apps, services, integrations, AI workflows, and analytics.
AI serving
Use real-time AI only when someone is waiting.
Real-timeSelf-assessment
Risk forms return results while the patient is waiting.
Real-time + batchSymptom extraction
Used during symptom flows and later to clean imported records.
Near real-timeVitals analysis
Alerts and trend summaries need low latency with clinical guardrails.
AsyncImaging workflows
Heavier image analysis is queued, reviewed, and governed before use.
Clinical governance
No clinical model should ship without proof.
Model cards, intended-use scope, validation results, interpretability evidence, subgroup performance, and drift monitoring make AI safer to operate.